1.. _pagemap:
2
3=============================
4Examining Process Page Tables
5=============================
6
7pagemap is a new (as of 2.6.25) set of interfaces in the kernel that allow
8userspace programs to examine the page tables and related information by
9reading files in ``/proc``.
10
11There are four components to pagemap:
12
13 * ``/proc/pid/pagemap``.  This file lets a userspace process find out which
14   physical frame each virtual page is mapped to.  It contains one 64-bit
15   value for each virtual page, containing the following data (from
16   ``fs/proc/task_mmu.c``, above pagemap_read):
17
18    * Bits 0-54  page frame number (PFN) if present
19    * Bits 0-4   swap type if swapped
20    * Bits 5-54  swap offset if swapped
21    * Bit  55    pte is soft-dirty (see
22      :ref:`Documentation/admin-guide/mm/soft-dirty.rst <soft_dirty>`)
23    * Bit  56    page exclusively mapped (since 4.2)
24    * Bits 57-60 zero
25    * Bit  61    page is file-page or shared-anon (since 3.5)
26    * Bit  62    page swapped
27    * Bit  63    page present
28
29   Since Linux 4.0 only users with the CAP_SYS_ADMIN capability can get PFNs.
30   In 4.0 and 4.1 opens by unprivileged fail with -EPERM.  Starting from
31   4.2 the PFN field is zeroed if the user does not have CAP_SYS_ADMIN.
32   Reason: information about PFNs helps in exploiting Rowhammer vulnerability.
33
34   If the page is not present but in swap, then the PFN contains an
35   encoding of the swap file number and the page's offset into the
36   swap. Unmapped pages return a null PFN. This allows determining
37   precisely which pages are mapped (or in swap) and comparing mapped
38   pages between processes.
39
40   Efficient users of this interface will use ``/proc/pid/maps`` to
41   determine which areas of memory are actually mapped and llseek to
42   skip over unmapped regions.
43
44 * ``/proc/kpagecount``.  This file contains a 64-bit count of the number of
45   times each page is mapped, indexed by PFN.
46
47 * ``/proc/kpageflags``.  This file contains a 64-bit set of flags for each
48   page, indexed by PFN.
49
50   The flags are (from ``fs/proc/page.c``, above kpageflags_read):
51
52    0. LOCKED
53    1. ERROR
54    2. REFERENCED
55    3. UPTODATE
56    4. DIRTY
57    5. LRU
58    6. ACTIVE
59    7. SLAB
60    8. WRITEBACK
61    9. RECLAIM
62    10. BUDDY
63    11. MMAP
64    12. ANON
65    13. SWAPCACHE
66    14. SWAPBACKED
67    15. COMPOUND_HEAD
68    16. COMPOUND_TAIL
69    17. HUGE
70    18. UNEVICTABLE
71    19. HWPOISON
72    20. NOPAGE
73    21. KSM
74    22. THP
75    23. BALLOON
76    24. ZERO_PAGE
77    25. IDLE
78
79 * ``/proc/kpagecgroup``.  This file contains a 64-bit inode number of the
80   memory cgroup each page is charged to, indexed by PFN. Only available when
81   CONFIG_MEMCG is set.
82
83Short descriptions to the page flags
84====================================
85
860 - LOCKED
87   page is being locked for exclusive access, e.g. by undergoing read/write IO
887 - SLAB
89   page is managed by the SLAB/SLOB/SLUB/SLQB kernel memory allocator
90   When compound page is used, SLUB/SLQB will only set this flag on the head
91   page; SLOB will not flag it at all.
9210 - BUDDY
93    a free memory block managed by the buddy system allocator
94    The buddy system organizes free memory in blocks of various orders.
95    An order N block has 2^N physically contiguous pages, with the BUDDY flag
96    set for and _only_ for the first page.
9715 - COMPOUND_HEAD
98    A compound page with order N consists of 2^N physically contiguous pages.
99    A compound page with order 2 takes the form of "HTTT", where H donates its
100    head page and T donates its tail page(s).  The major consumers of compound
101    pages are hugeTLB pages
102    (:ref:`Documentation/admin-guide/mm/hugetlbpage.rst <hugetlbpage>`),
103    the SLUB etc.  memory allocators and various device drivers.
104    However in this interface, only huge/giga pages are made visible
105    to end users.
10616 - COMPOUND_TAIL
107    A compound page tail (see description above).
10817 - HUGE
109    this is an integral part of a HugeTLB page
11019 - HWPOISON
111    hardware detected memory corruption on this page: don't touch the data!
11220 - NOPAGE
113    no page frame exists at the requested address
11421 - KSM
115    identical memory pages dynamically shared between one or more processes
11622 - THP
117    contiguous pages which construct transparent hugepages
11823 - BALLOON
119    balloon compaction page
12024 - ZERO_PAGE
121    zero page for pfn_zero or huge_zero page
12225 - IDLE
123    page has not been accessed since it was marked idle (see
124    :ref:`Documentation/admin-guide/mm/idle_page_tracking.rst <idle_page_tracking>`).
125    Note that this flag may be stale in case the page was accessed via
126    a PTE. To make sure the flag is up-to-date one has to read
127    ``/sys/kernel/mm/page_idle/bitmap`` first.
128
129IO related page flags
130---------------------
131
1321 - ERROR
133   IO error occurred
1343 - UPTODATE
135   page has up-to-date data
136   ie. for file backed page: (in-memory data revision >= on-disk one)
1374 - DIRTY
138   page has been written to, hence contains new data
139   i.e. for file backed page: (in-memory data revision >  on-disk one)
1408 - WRITEBACK
141   page is being synced to disk
142
143LRU related page flags
144----------------------
145
1465 - LRU
147   page is in one of the LRU lists
1486 - ACTIVE
149   page is in the active LRU list
15018 - UNEVICTABLE
151   page is in the unevictable (non-)LRU list It is somehow pinned and
152   not a candidate for LRU page reclaims, e.g. ramfs pages,
153   shmctl(SHM_LOCK) and mlock() memory segments
1542 - REFERENCED
155   page has been referenced since last LRU list enqueue/requeue
1569 - RECLAIM
157   page will be reclaimed soon after its pageout IO completed
15811 - MMAP
159   a memory mapped page
16012 - ANON
161   a memory mapped page that is not part of a file
16213 - SWAPCACHE
163   page is mapped to swap space, i.e. has an associated swap entry
16414 - SWAPBACKED
165   page is backed by swap/RAM
166
167The page-types tool in the tools/vm directory can be used to query the
168above flags.
169
170Using pagemap to do something useful
171====================================
172
173The general procedure for using pagemap to find out about a process' memory
174usage goes like this:
175
176 1. Read ``/proc/pid/maps`` to determine which parts of the memory space are
177    mapped to what.
178 2. Select the maps you are interested in -- all of them, or a particular
179    library, or the stack or the heap, etc.
180 3. Open ``/proc/pid/pagemap`` and seek to the pages you would like to examine.
181 4. Read a u64 for each page from pagemap.
182 5. Open ``/proc/kpagecount`` and/or ``/proc/kpageflags``.  For each PFN you
183    just read, seek to that entry in the file, and read the data you want.
184
185For example, to find the "unique set size" (USS), which is the amount of
186memory that a process is using that is not shared with any other process,
187you can go through every map in the process, find the PFNs, look those up
188in kpagecount, and tally up the number of pages that are only referenced
189once.
190
191Other notes
192===========
193
194Reading from any of the files will return -EINVAL if you are not starting
195the read on an 8-byte boundary (e.g., if you sought an odd number of bytes
196into the file), or if the size of the read is not a multiple of 8 bytes.
197
198Before Linux 3.11 pagemap bits 55-60 were used for "page-shift" (which is
199always 12 at most architectures). Since Linux 3.11 their meaning changes
200after first clear of soft-dirty bits. Since Linux 4.2 they are used for
201flags unconditionally.
202